skmelendez · 2026-02-25T18:26:12Z

Description

Three-part PR that adds cross-platform runtime support, improves the UI scaling system, and establishes a full test suite.

Cross-platform support — Extracted all platform-specific logic (process listing, TCP connections, file handles, process control) into a platform/ abstraction layer with dedicated modules for darwin, linux, and win32. Refactored process-scanner, process-utils, file-watcher, and network-monitor to use the platform layer instead of hardcoded Windows calls. The app now runs on macOS and Linux in addition to Windows.

Unified UI scaling — Replaced the font-only --aegis-font-scale with a unified --aegis-ui-scale system. One slider now controls text, spacing, padding, gaps, and container sizing together. Added ~20 scaling tokens to tokens.css and converted ~106 hardcoded px values across 24+ Svelte components. Also fixed radar high-contrast rendering. Backward-compatible store migration from the old localStorage key.

Test suite — Added 386 tests across 21 test files using vitest + v8 coverage. Achieves 86% statement and 88% line coverage. Tests cover anomaly scoring, AI analysis API handling, CSV/HTML export escaping, platform process parsing, tray icon threat colors, IPC handler registration, and more.

Multiple modules hit 100% coverage (ai-analysis, exports, threat-report, risk-scoring, constants).

Added DI test hooks (_resetForTest, _setDepsForTest, _setPlatformForTest) to 8 source modules for clean mocking. Added CI test job, npm test scripts, and vitest/coverage devDependencies.

Type of change

New feature
Bug fix

Testing

All 386 tests pass via npx vitest run
Coverage verified via npx vitest run --coverage (86% stmts, 88% lines)
Tested locally on macOS with npm start — app launches, scans processes, monitors files, network panel functional

Checklist

Follows code style guidelines (CommonJS main, Svelte 5 renderer)
No console.log in production code
Files under 200 lines
I have tested this locally with npm start
This code has been reviewed by a human (not solely AI-generated)

…n, but does run on mac, will speak to repo owner about getting on discord to see which features are needed

Implement a cohesive interface scaling system that replaces the font-only --aegis-font-scale with a unified --aegis-ui-scale variable. Previously, increasing the scale slider only affected font sizes while layout spacing remained fixed, creating visual imbalance. Now one slider controls text, spacing, padding, gaps, and container sizing together. Scaling tokens added to tokens.css: - 10 spacing tokens (--aegis-space-1 through --aegis-space-10, 2px–24px) - 6 sizing tokens for structural elements (header, footer, timeline, panel min-widths, modal min-width) - 4 column-width tokens for tabular layouts (time, action, agent, state) - All tokens use calc(Npx * var(--aegis-ui-scale)) pattern Converted ~106 hardcoded px values across 24+ Svelte components to use the new tokens. One-off constraints (e.g. max-height on scrollable areas) use inline calc() with --aegis-ui-scale. Left as literal px (intentionally not scaled): - 1px borders/dividers, border-radius (uses shape tokens), box-shadow - Canvas/SVG internals (devicePixelRatio handles canvas scaling) - Media query breakpoints, animation transforms, toggle switch internals Store changes (theme.js): - Renamed fontScale → uiScale, setFontScale → setUiScale - localStorage key migrated from aegis-font-scale to aegis-ui-scale with automatic fallback read of the old key - Backward-compatible re-exports preserved Config changes (config-manager.js): - Default settings key renamed fontScale → uiScale Radar high-contrast theme fix: - Added missing CSS custom property overrides for dark-hc and light-hc themes (previously only dark and light had radar-specific RGB values) - Fixed isLight detection to include light-hc variant - Added isHC flag that boosts alpha values for rings, crosshairs, sweep trail, agent dots, and labels in high-contrast modes - Dark HC: brighter line/label RGB, more vivid sweep color - Light HC: darker line/label RGB, stronger sweep contrast

Add 386 tests across 21 test files, bringing statement coverage from ~44% file coverage to 86% overall. All new tests validate real logic with no placeholders or stubs. New test files: - anomaly-detector (25 tests) — scoring weights, deviation warnings, dedup - ai-analysis (22 tests) — API calls, JSON extraction, error handling - exports (12 tests) — CSV escaping, JSON export, HTML report generation - tray-icon (20 tests) — PNG generation, threat colors, notifications - ipc-handlers (17 tests) — channel registration, handler delegation - platform/darwin (15 tests) — ps parsing, .app bundle extraction - platform/win32 (40 tests) — tasklist parsing, PowerShell integration - platform/index (3 tests) — platform dispatch verification - threat-report (9 tests) — HTML generation, XSS escaping, risk colors Coverage highlights: - ai-analysis.js: 100% statements/lines - anomaly-detector.js: 99%/100% - exports.js: 100%/100% - threat-report.js: 100%/100% - win32.js: 99%/99% Updated vitest.config.js coverage include list to track all new modules.

Add dependency injection test helpers (_resetForTest, _setDepsForTest, _setPlatformForTest, _setExecFileForTest) to 8 source modules so tests can mock platform calls without brittle module patching. - baselines.js: _setBaselinesPathForTest - config-manager.js: _setSettingsPathForTest - file-watcher.js: _setDepsForTest, _resetForTest, export isSelfAccess/handleWatcherEvent - network-monitor.js: _setDepsForTest, _resetForTest, export isPrivateIp/resolveIp - process-scanner.js: _setPlatformForTest, _resetForTest - process-utils.js: _setPlatformForTest, _resetForTest - platform/posix-shared.js: _setExecFileForTest - platform/linux.js: _setExecFileForTest, export parseSsOutput Also adds: - CI: test job in .github/workflows/ci.yml - package.json: test/test:watch/test:coverage scripts, vitest + @vitest/coverage-v8 devDeps - README: testing section with commands

…ne follow - Add Ctrl/Cmd+Shift+T global shortcut to cycle through all four themes - Theme footer with CSS tokens so it's visible in light/HC themes instead of hardcoded dark background - Show active agent count in system tray tooltip - Fix audit log stats not displaying: backend returned mismatched field names (todayEntries/totalFiles/recordingSince) vs what the UI expected (totalEntries/totalSize/firstEntry/lastEntry); now returns correct fields with actual entry counts, file sizes, and timestamp ranges - Audit log size now shows current log and total: "12.3 KB (1.2 MB)" with support for B/KB/MB/GB/TB - Center radar in shield layout using 3-column grid with agents panel to the right - Timeline scrubber defaults to right edge and auto-follows as new events arrive; user can drag away to browse history, drag back to right to re-enable following - Persist timeline zoom level in settings across restarts

Timeline lazy-loads historical audit log entries when the user drags the scrubber to the left edge. The scrubber is always freely draggable; it defaults to the right edge (following live events) and only triggers a history fetch when it physically reaches position 0. Each fetch loads 25 entries at a time using a performant reverse file reader (4KB chunks from EOF, stops as soon as the batch is full, skips files by date). After each load the viewport scroll compensates for the expanded time range so the view stays stable instead of jumping. - Add audit-logger getEntriesBefore() with chunked reverse file reading - Add get-audit-entries-before IPC handler + preload bridge - Decouple thumb position from scrollLeft so scrubber is always movable - Merge historical + live events with timestamp dedup, cap at 500 - Show active agent count in tray context menu, rebuilt on each scan

Fix/UI polish

Timeline lazy-loads historical audit log entries when the user drags the scrubber to the left edge. The scrubber is always freely draggable; it defaults to the right edge (following live events) and only triggers a history fetch when it physically reaches position 0. Each fetch loads 25 entries at a time using a performant reverse file reader (4KB chunks from EOF, stops as soon as the batch is full, skips files by date). After each load the viewport scroll compensates for the expanded time range so the view stays stable instead of jumping. - Add audit-logger getEntriesBefore() with chunked reverse file reading - Add get-audit-entries-before IPC handler + preload bridge - Decouple thumb position from scrollLeft so scrubber is always movable - Merge historical + live events with timestamp dedup, cap at 500 - Show active agent count in tray context menu, rebuilt on each scan

@skmelendez

Fixes identified during maintainer code review of #37 (cross-platform support by @skmelendez). No business logic changes — only safety guards, DI consistency, and test portability. Platform layer: - linux.js: fix bare execFile → _execFile in getParentProcessMap ps-fallback (was bypassing DI hook, breaking test mockability on Linux without /proc) - darwin.js: add _setExecFileForTest DI hook for test mockability (consistent with linux.js and posix-shared.js pattern) - win32.js: add PID integer validation to killProcess, suspendProcess, resumeProcess (defence-in-depth, matches existing getFileHandles guard) Tests: - file-watcher.test.js: skip 3 darwin-specific shouldIgnore tests on non-macOS platforms (patterns from darwin.js not loaded on Windows) - ipc-handlers.test.js: case-insensitive directory name assertion (Windows is case-preserving: AEGIS vs Aegis) Reviewed-by: @antropos17 Co-authored-by: @skmelendez

antropos17

Great work on cross-platform support @skmelendez! This is a massive contribution — platform abstraction layer, 386 tests, UI scaling system.

I've done a full code review and identified a few critical fixes needed (linux.js DI hook, win32 PID validation, test portability). I'll push those as a follow-up PR referencing this one.

Merging — thank you for the contribution! 🎉

@skmelendez

Fixes identified during maintainer code review of #37 (cross-platform support by @skmelendez). No business logic changes -- only safety guards, DI consistency, and test portability. Platform layer: - linux.js: fix bare execFile -> _execFile in getParentProcessMap ps-fallback (was bypassing DI hook, breaking test mockability on Linux without /proc) - darwin.js: add _setExecFileForTest DI hook for test mockability (consistent with linux.js and posix-shared.js pattern) - win32.js: add PID integer validation to killProcess, suspendProcess, resumeProcess (defence-in-depth, matches existing getFileHandles guard) Tests: - file-watcher.test.js: skip 4 darwin/linux-specific shouldIgnore tests on non-matching platforms (patterns from darwin.js/linux.js not loaded on Windows) - ipc-handlers.test.js: case-insensitive directory name assertion (Windows is case-preserving: AEGIS vs Aegis) Reviewed-by: @antropos17 Co-authored-by: @skmelendez

fix: critical issues from PR #37 code review

Replace split('/') with split(/[/\]/) to handle both Unix and Windows paths. Fixes path display truncation on Windows where backslash paths were not split correctly. Resolves: PR #37 code review HIGH issue #1 Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

Wrap onToggleTheme listener in $effect with cleanup return. Make preload.js onToggleTheme return an unsubscribe function. Prevents duplicate listeners accumulating during hot module replacement. Resolves: PR #37 code review HIGH issue #4 Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

Only fall back to lsof when ss fails with error, not on empty results. An agent with zero TCP connections is normal — no need to retry with lsof. Updated test to expect correct behavior (empty result, no lsof fallback). Resolves: PR #37 code review HIGH issue #7 Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

Prevents monkey-patched module loader from leaking between test suites if Vitest isolation settings change. Resolves: PR #37 code review HIGH issue #8 Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

* fix: use cross-platform path separator in AgentCard Replace split('/') with split(/[/\]/) to handle both Unix and Windows paths. Fixes path display truncation on Windows where backslash paths were not split correctly. Resolves: PR #37 code review HIGH issue #1 Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com> * fix: cleanup IPC listener to prevent HMR leak in App.svelte Wrap onToggleTheme listener in $effect with cleanup return. Make preload.js onToggleTheme return an unsubscribe function. Prevents duplicate listeners accumulating during hot module replacement. Resolves: PR #37 code review HIGH issue #4 Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com> * fix: correct ss→lsof fallback logic in linux getRawTcpConnections Only fall back to lsof when ss fails with error, not on empty results. An agent with zero TCP connections is normal — no need to retry with lsof. Updated test to expect correct behavior (empty result, no lsof fallback). Resolves: PR #37 code review HIGH issue #7 Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com> * fix: restore Module._load in afterAll for 6 test files Prevents monkey-patched module loader from leaking between test suites if Vitest isolation settings change. Resolves: PR #37 code review HIGH issue #8 Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com> --------- Co-authored-by: Claude Opus 4.6 <noreply@anthropic.com>

darwin.js and linux.js had identical ps output parsing logic. Shared function handles baseline parsing; darwin adds .app bundle name extraction as post-processing. Resolves: PR #37 code review HIGH issue #5 Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

Ensures test mocks match real return shapes from main.js, audit-logger.js, and logger.js getStats() functions. Improves test reliability and documentation of API contracts. Resolves: PR #37 code review HIGH issue #6 Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

* refactor: extract parsePsOutput into posix-shared for DRY darwin.js and linux.js had identical ps output parsing logic. Shared function handles baseline parsing; darwin adds .app bundle name extraction as post-processing. Resolves: PR #37 code review HIGH issue #5 Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com> * test: align getStats mock shapes with actual API contracts Ensures test mocks match real return shapes from main.js, audit-logger.js, and logger.js getStats() functions. Improves test reliability and documentation of API contracts. Resolves: PR #37 code review HIGH issue #6 Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com> --------- Co-authored-by: Claude Opus 4.6 <noreply@anthropic.com>

…omprehensive test suite (antropos17#37) * Updated claude for base run on linux/mac, will need further validation, but does run on mac, will speak to repo owner about getting on discord to see which features are needed * feat: unified UI scaling system + radar high-contrast fix Implement a cohesive interface scaling system that replaces the font-only --aegis-font-scale with a unified --aegis-ui-scale variable. Previously, increasing the scale slider only affected font sizes while layout spacing remained fixed, creating visual imbalance. Now one slider controls text, spacing, padding, gaps, and container sizing together. Scaling tokens added to tokens.css: - 10 spacing tokens (--aegis-space-1 through --aegis-space-10, 2px–24px) - 6 sizing tokens for structural elements (header, footer, timeline, panel min-widths, modal min-width) - 4 column-width tokens for tabular layouts (time, action, agent, state) - All tokens use calc(Npx * var(--aegis-ui-scale)) pattern Converted ~106 hardcoded px values across 24+ Svelte components to use the new tokens. One-off constraints (e.g. max-height on scrollable areas) use inline calc() with --aegis-ui-scale. Left as literal px (intentionally not scaled): - 1px borders/dividers, border-radius (uses shape tokens), box-shadow - Canvas/SVG internals (devicePixelRatio handles canvas scaling) - Media query breakpoints, animation transforms, toggle switch internals Store changes (theme.js): - Renamed fontScale → uiScale, setFontScale → setUiScale - localStorage key migrated from aegis-font-scale to aegis-ui-scale with automatic fallback read of the old key - Backward-compatible re-exports preserved Config changes (config-manager.js): - Default settings key renamed fontScale → uiScale Radar high-contrast theme fix: - Added missing CSS custom property overrides for dark-hc and light-hc themes (previously only dark and light had radar-specific RGB values) - Fixed isLight detection to include light-hc variant - Added isHC flag that boosts alpha values for rings, crosshairs, sweep trail, agent dots, and labels in high-contrast modes - Dark HC: brighter line/label RGB, more vivid sweep color - Light HC: darker line/label RGB, stronger sweep contrast * test: add comprehensive test coverage for 9 previously untested modules Add 386 tests across 21 test files, bringing statement coverage from ~44% file coverage to 86% overall. All new tests validate real logic with no placeholders or stubs. New test files: - anomaly-detector (25 tests) — scoring weights, deviation warnings, dedup - ai-analysis (22 tests) — API calls, JSON extraction, error handling - exports (12 tests) — CSV escaping, JSON export, HTML report generation - tray-icon (20 tests) — PNG generation, threat colors, notifications - ipc-handlers (17 tests) — channel registration, handler delegation - platform/darwin (15 tests) — ps parsing, .app bundle extraction - platform/win32 (40 tests) — tasklist parsing, PowerShell integration - platform/index (3 tests) — platform dispatch verification - threat-report (9 tests) — HTML generation, XSS escaping, risk colors Coverage highlights: - ai-analysis.js: 100% statements/lines - anomaly-detector.js: 99%/100% - exports.js: 100%/100% - threat-report.js: 100%/100% - win32.js: 99%/99% Updated vitest.config.js coverage include list to track all new modules. * chore: add test infrastructure, DI hooks, CI test job, and devDeps Add dependency injection test helpers (_resetForTest, _setDepsForTest, _setPlatformForTest, _setExecFileForTest) to 8 source modules so tests can mock platform calls without brittle module patching. - baselines.js: _setBaselinesPathForTest - config-manager.js: _setSettingsPathForTest - file-watcher.js: _setDepsForTest, _resetForTest, export isSelfAccess/handleWatcherEvent - network-monitor.js: _setDepsForTest, _resetForTest, export isPrivateIp/resolveIp - process-scanner.js: _setPlatformForTest, _resetForTest - process-utils.js: _setPlatformForTest, _resetForTest - platform/posix-shared.js: _setExecFileForTest - platform/linux.js: _setExecFileForTest, export parseSsOutput Also adds: - CI: test job in .github/workflows/ci.yml - package.json: test/test:watch/test:coverage scripts, vitest + @vitest/coverage-v8 devDeps - README: testing section with commands * chore: add coverage/ to .gitignore * feat: UI polish — theme shortcut, footer theming, audit stats, timeline follow - Add Ctrl/Cmd+Shift+T global shortcut to cycle through all four themes - Theme footer with CSS tokens so it's visible in light/HC themes instead of hardcoded dark background - Show active agent count in system tray tooltip - Fix audit log stats not displaying: backend returned mismatched field names (todayEntries/totalFiles/recordingSince) vs what the UI expected (totalEntries/totalSize/firstEntry/lastEntry); now returns correct fields with actual entry counts, file sizes, and timestamp ranges - Audit log size now shows current log and total: "12.3 KB (1.2 MB)" with support for B/KB/MB/GB/TB - Center radar in shield layout using 3-column grid with agents panel to the right - Timeline scrubber defaults to right edge and auto-follows as new events arrive; user can drag away to browse history, drag back to right to re-enable following - Persist timeline zoom level in settings across restarts * feat: timeline history scroll-back, tray menu agent count Timeline lazy-loads historical audit log entries when the user drags the scrubber to the left edge. The scrubber is always freely draggable; it defaults to the right edge (following live events) and only triggers a history fetch when it physically reaches position 0. Each fetch loads 25 entries at a time using a performant reverse file reader (4KB chunks from EOF, stops as soon as the batch is full, skips files by date). After each load the viewport scroll compensates for the expanded time range so the view stays stable instead of jumping. - Add audit-logger getEntriesBefore() with chunked reverse file reading - Add get-audit-entries-before IPC handler + preload bridge - Decouple thumb position from scrollLeft so scrubber is always movable - Merge historical + live events with timestamp dedup, cap at 500 - Show active agent count in tray context menu, rebuilt on each scan * feat: timeline history scroll-back, tray menu agent count Timeline lazy-loads historical audit log entries when the user drags the scrubber to the left edge. The scrubber is always freely draggable; it defaults to the right edge (following live events) and only triggers a history fetch when it physically reaches position 0. Each fetch loads 25 entries at a time using a performant reverse file reader (4KB chunks from EOF, stops as soon as the batch is full, skips files by date). After each load the viewport scroll compensates for the expanded time range so the view stays stable instead of jumping. - Add audit-logger getEntriesBefore() with chunked reverse file reading - Add get-audit-entries-before IPC handler + preload bridge - Decouple thumb position from scrollLeft so scrubber is always movable - Merge historical + live events with timestamp dedup, cap at 500 - Show active agent count in tray context menu, rebuilt on each scan

@skmelendez

Fixes identified during maintainer code review of antropos17#37 (cross-platform support by @skmelendez). No business logic changes -- only safety guards, DI consistency, and test portability. Platform layer: - linux.js: fix bare execFile -> _execFile in getParentProcessMap ps-fallback (was bypassing DI hook, breaking test mockability on Linux without /proc) - darwin.js: add _setExecFileForTest DI hook for test mockability (consistent with linux.js and posix-shared.js pattern) - win32.js: add PID integer validation to killProcess, suspendProcess, resumeProcess (defence-in-depth, matches existing getFileHandles guard) Tests: - file-watcher.test.js: skip 4 darwin/linux-specific shouldIgnore tests on non-matching platforms (patterns from darwin.js/linux.js not loaded on Windows) - ipc-handlers.test.js: case-insensitive directory name assertion (Windows is case-preserving: AEGIS vs Aegis) Reviewed-by: @antropos17 Co-authored-by: @skmelendez

fix: critical issues from PR antropos17#37 code review

…s17#41) * fix: use cross-platform path separator in AgentCard Replace split('/') with split(/[/\]/) to handle both Unix and Windows paths. Fixes path display truncation on Windows where backslash paths were not split correctly. Resolves: PR antropos17#37 code review HIGH issue #1 Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com> * fix: cleanup IPC listener to prevent HMR E3A7 leak in App.svelte Wrap onToggleTheme listener in $effect with cleanup return. Make preload.js onToggleTheme return an unsubscribe function. Prevents duplicate listeners accumulating during hot module replacement. Resolves: PR antropos17#37 code review HIGH issue antropos17#4 Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com> * fix: correct ss→lsof fallback logic in linux getRawTcpConnections Only fall back to lsof when ss fails with error, not on empty results. An agent with zero TCP connections is normal — no need to retry with lsof. Updated test to expect correct behavior (empty result, no lsof fallback). Resolves: PR antropos17#37 code review HIGH issue antropos17#7 Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com> * fix: restore Module._load in afterAll for 6 test files Prevents monkey-patched module loader from leaking between test suites if Vitest isolation settings change. Resolves: PR antropos17#37 code review HIGH issue antropos17#8 Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com> --------- Co-authored-by: Claude Opus 4.6 <noreply@anthropic.com>

* refactor: extract parsePsOutput into posix-shared for DRY darwin.js and linux.js had identical ps output parsing logic. Shared function handles baseline parsing; darwin adds .app bundle name extraction as post-processing. Resolves: PR antropos17#37 code review HIGH issue antropos17#5 Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com> * test: align getStats mock shapes with actual API contracts Ensures test mocks match real return shapes from main.js, audit-logger.js, and logger.js getStats() functions. Improves test reliability and documentation of API contracts. Resolves: PR antropos17#37 code review HIGH issue antropos17#6 Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com> --------- Co-authored-by: Claude Opus 4.6 <noreply@anthropic.com>

…ess entry (#311)

skmelendez added 8 commits February 25, 2026 07:51

Updated claude for base run on linux/mac, will need further validatio…

225f84b

…n, but does run on mac, will speak to repo owner about getting on discord to see which features are needed

chore: add coverage/ to .gitignore

0194ca3

Merge pull request #1 from skmelendez/fix/ui-polish

cadefc7

Fix/UI polish

This was referenced Feb 25, 2026

🤖 Add active agent count to system tray tooltip #36

Closed

🤖 Add theme toggle keyboard shortcut (Ctrl+Shift+T) #35

Closed

Add dark/light theme toggle persistence #29

Closed

Add test suite (Vitest + Playwright) #11

Closed

skmelendez mentioned this pull request Feb 25, 2026

High Contrast Light/Dark mode themes #38

Closed

antropos17 approved these changes Feb 28, 2026

View reviewed changes

antropos17 merged commit 4abfe6f into antropos17:master Feb 28, 2026

antropos17 mentioned this pull request Feb 28, 2026

fix: critical issues from PR #37 code review #40

Merged

antropos17 added a commit that referenced this pull request Feb 28, 2026

Merge pull request #40 from antropos17/fix/pr37-critical-fixes

f43a7e9

fix: critical issues from PR #37 code review

antropos17 mentioned this pull request Feb 28, 2026

fix: resolve 4 HIGH issues from PR #37 code review #41

Merged

antropos17 mentioned this pull request Feb 28, 2026

chore: format entire codebase with Prettier #42

Merged

antropos17 mentioned this pull request Feb 28, 2026

refactor: DRY platform code + align test API contracts #43

Merged

github-actions Bot mentioned this pull request Mar 2, 2026

chore(master): release aegis 0.3.1-alpha #59

Merged

ElshadHu pushed a commit to ElshadHu/Aegis that referenced this pull request Apr 9, 2026

Merge pull request antropos17#40 from antropos17/fix/pr37-critical-fixes

da97ce3

fix: critical issues from PR antropos17#37 code review

antropos17 mentioned this pull request Aug 25, 2026

docs(memory-bank): record ai-mistakes #37-#39 and the #308-#310 progress entry #311

Merged

antropos17 added a commit that referenced this pull request Aug 25, 2026

docs(memory-bank): record ai-mistakes #37-#39 and the #308-#310 progr…

b5db377

…ess entry (#311)

github-actions Bot mentioned this pull request Aug 25, 2026

chore(master): release aegis 0.14.0-alpha #287

Open

antropos17 mentioned this pull request Aug 25, 2026

docs(memory-bank): ai-mistakes #41 — git worktree remove deletes through a node_modules junction #326

Merged

9 tasks

Navigation Menu

feat: cross-platform support (macOS/Linux), unified UI scaling, and comprehensive test suite - #37